2017/12/27 by Kostas Hatalis, Hatalis, Kostas, Shalinee Kishore +1
Computer Science · Decision Sciences · Engineering · #Energy Load and Power Forecasting #FOS: Computer and information sciences #Forecasting Techniques and Applications #Image and Signal Denoising Methods #Machine Learning (stat.ML) #Neural Networks and Applications #Stock Market Forecasting Methods
paper · pdf · doi:10.48550/arxiv.1712.09641
openalex publication_date 2017/12/27 · openalex created_date 2022/09/02 · openalex updated_date 2026/07/28
Point forecasting of univariate time series is a challenging problem with\nextensive work having been conducted. However, nonparametric probabilistic\nforecasting of time series, such as in the form of quantiles or prediction\nintervals is an even more challenging problem. In an effort to expand the\npossible forecasting paradigms we devise and explore an extrapolation-based\napproach that has not been applied before for probabilistic forecasting. We\npresent a novel quantile Fourier neural network is for nonparametric\nprobabilistic forecasting of univariate time series. Multi-step predictions are\nprovided in the form of composite quantiles using time as the only input to the\nmodel. This effectively is a form of extrapolation based nonlinear quantile\nregression applied for forecasting. Experiments are conducted on eight real\nworld datasets that demonstrate a variety of periodic and aperiodic patterns.\nNine naive and advanced methods are used as benchmarks including quantile\nregression neural network, support vector quantile regression, SARIMA, and\nexponential smoothing. The obtained empirical results validate the\neffectiveness of the proposed method in providing high quality and accurate\nprobabilistic predictions.\n